Data.Vector.Unboxed.Deriving

Writing Unbox instances for new data types is tedious and formulaic. More often than not, there is a straightforward mapping of the new type onto some existing one already imbued with an Unbox instance. The example from the vector package represents Complex a as pairs (a, a). (See http://hackage.haskell.org/packages/archive/vector/latest/doc/html/Data-Vector-Unboxed.html.) Using derivingUnbox, we can define the same instances much more succinctly:

derivingUnbox "Complex"
    [d| instance (Unbox a) => Unbox' (Complex a) (a, a) |]
    [| \ (r :+ i) -> (r, i) |]
    [| \ (r, i) -> r :+ i |]

Requires the MultiParamTypeClasses, TemplateHaskell and TypeFamilies LANGUAGE extensions.

The dummy Unbox' class provides a convenient way to pass in the source and representation types, along with any requisite constraints.

Documentation

class Unbox' src rep

A dummy class for passing arguments to derivingUnbox.

derivingUnbox

Arguments

:: String

Unique constructor suffix for the MVector and Vector data families

-> DecsQ

Quotation of the form [d| instance ctxt => Unbox' src rep |]

-> ExpQ

Quotation of an expression of type src -> rep

-> ExpQ

Quotation of an expression of type rep -> src

-> DecsQ

Declarations to be spliced for the derived Unbox instance

Let's consider a more complex example: suppose we want an Unbox instance for Maybe a. We can encode this using the pair (Bool, a), with the boolean indicating whether we have Nothing or Just something. This encoding requires a dummy value in the Nothing case, necessitating an additional Default (see the data-default package) constraint. Thus:

derivingUnbox "Maybe"
    [d| instance (Default a, Unbox a) => Unbox' (Maybe a) (Bool, a) |]
    [| maybe (False, def) (\ x -> (True, x)) |]
    [| \ (b, x) -> if b then Just x else Nothing |]